MGTANet: Encoding Sequential LiDAR Points Using Long Short-Term Motion-Guided Temporal Attention for 3D Object Detection.
Junho KohJunhyung LeeYoungwoo LeeJaekyum KimJun Won ChoiPublished in: AAAI (2023)
Keyphrases
- short term
- d objects
- long term
- motion prediction
- viewpoint
- pose parameters
- physically plausible
- space time
- viewing direction
- lidar data
- multi view
- geometric invariants
- surface points
- short term and long term
- pose estimation
- object recognition
- range data
- feature points
- multiple views
- shape descriptors
- object contours
- surface patches
- short and long term
- three dimensional
- motion estimation
- object detection
- point distribution
- three dimensional objects
- minimum bounding
- load forecasting
- d mesh
- image sequences
- pose normalization
- object classes
- point sets
- forecasting model
- cad models
- optical flow
- medium term
- long term memory
- camera motion
- point cloud
- computer vision
- topological information
- moving objects
- surface properties
- least squares
- rigid motion
- control points